Reimagining the GPU for real-time interactive AI
We are engineers, we are dreamers. Together, we built GPUs that power billions of computers. We love pixels, matrices, neural networks, and transistors. The silicon we built made your cell phones into tera-scale supercomputers and ushered in exa-scale computing in the data centers. Our individual journeys transported us inside cultures like AMD, Apple, Intel, Nvidia and several startups. Together, the dreams we shared changed the world.
The first floating-point programmable GPU architecture we launched in 2002 is still the dominant DNA that evolved GPUs into the era of AI.
We got the band back together in 2023 with a new mission: planting a new GPU architecture seed that will take us into the 2040s. But sometimes, to go forward we first must go backward. So we went back to principles: software and physics, Python and memory, generality and performance.
We know moon shots are relentless incremental jumps over many roof tops. We also know that the next breakthrough in intelligence will require one necessary ingredient: human ingenuity.
We have offices in California, Canada, and in India at Hyderabad.